# GNSS-R Soil Moisture Retrieval Challenge

**Advancing Global Soil Moisture Monitoring with Spaceborne GNSS Reflectometry**

## Competition Summary
This competition leverages publicly available GNSS-Reflectometry (GNSS-R) data from NASA's CYGNSS (Cyclone Global Navigation Satellite System) constellation (2017–2024) to push the boundaries of high-accuracy soil moisture retrieval using reflected satellite signals.

Participants will predict **volumetric surface soil moisture (0–5 cm depth, in cm³/cm³)** from GNSS-R reflectivity measurements and auxiliary features, addressing real-world challenges like vegetation cover, surface roughness, topography, and radio-frequency interference.

The dataset includes over 50 million high-quality samples across global land surfaces, with ground truth co-located from the International Soil Moisture Network (ISMN) in-situ stations and/or SMAP/SMOS satellite products.

## Task
- **Type**: Regression
- Predict soil moisture for each row in the test set.
- Submission format: CSV with columns `id` (or row index) and `soil_moisture`.

## Why Participate?
GNSS-R offers unique advantages: all-weather observations, high spatio-temporal density (~1–10 km daily in mid/low latitudes), and low-cost scalability compared to traditional microwave missions (e.g., SMAP at ~36 km).

Winning models could enable operational global soil moisture products, benefiting:
- Agricultural drought monitoring
- Hydrological forecasting
- Climate modeling and land-atmosphere interactions

## Dataset Highlights
- **Features**: Reflectivity (Γ), incidence/azimuth angles, coherence, surface roughness, vegetation opacity (τ), NDVI, elevation (DEM), land cover, time-series statistics, etc.
- **Scale**: ~45M training rows, ~5–10M test rows
- **Challenges**: Heterogeneous surfaces, seasonal variations, frozen soil, RFI

Join us to innovate in remote sensing and machine learning for Earth observation!

**Acknowledgments**  
Data provided by NASA CYGNSS mission (PO.DAAC), ISMN, and SMAP/SMOS teams.